Databricks Machine Learning Associate Question 251
Select 2A data scientist is evaluating a binary classification model that predicts whether a customer will churn. Given that false negatives (predicting a customer will not churn when they actually will) are more costly than false positives, which evaluation metric(s) would be most appropriate to focus on in this scenario?
- A
Recall
- B
Precision
- C
F1-Score
- D
Accuracy
Show answer and explanation
Correct answers: A, C
Explanation
When false negatives are more costly than false positives, recall is a critical metric because it directly measures the ability to identify all positive instances. F1-Score is also important as it balances recall and precision, providing a more comprehensive evaluation in cases where the costs of false negatives and false positives differ. Accuracy and precision are less relevant in this context due to their inability to effectively capture the importance of minimizing false negatives.
- A. Correct.
Recall is the proportion of true positives correctly identified out of all actual positives. Since false negatives are more costly in this scenario, recall is particularly important because it measures the model's ability to identify all positive cases.
- B. Incorrect.
Precision measures the proportion of true positives out of all predicted positives. While precision is important in some contexts, it is less relevant here because the main concern is minimizing false negatives, not false positives.
- C. Correct.
F1-Score is the harmonic mean of precision and recall. It is useful when there is an imbalance between false positives and false negatives, and it balances recall and precision. In this scenario, F1-Score complements recall as a key metric.
- D. Incorrect.
Accuracy measures the proportion of correctly classified instances out of all instances. However, it is not suitable in this case because it does not account for the imbalance between false positives and false negatives or the relative cost of these errors.